© 2017 de Bruyn et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
The study design, protocols and research instruments for this program were approved by the National Institute for Medical Research ethics committee (NIMR/HQ/R.8a/Vol.IX/1690) in Tanzania, The University of Sydney Human Research Ethics Committee (2014/209) and The University of Sydney Animal Ethics Committee (2013/6065). All participants provided informed consent prior to participating in the study, with assurance of confidentiality, anonymity, voluntary participation and no adverse effects in case of refusal. This longitudinal study comprises eight rural villages in two adjoining wards of Manyoni District in the semi-arid central zone of Tanzania, and represents a subset of a cluster randomised controlled trial evaluating the impact of village poultry vaccination programs and strategic improvements to crop systems on levels of chronic undernutrition in children [18]. Study sites were selected based on recommendations of in-country partners, according to levels of child stunting and the absence of existing human nutrition or poultry vaccination programs. Within the district, a unimodal pattern of rainfall is expected between November and April, with long-term data indicating mean annual rainfall of 624 mm (SD = 179 mm) and a mean number of rain days of 49 (SD = 15) [19]. A ward census was conducted by the project in April 2014 in Sanza Ward and October 2014 in Majiri Ward, with enumerators registering details of all household members, ownership of chickens and interest in chicken-keeping. Eligibility criteria for inclusion in the longitudinal study included the presence of a child under 24 months of age, and either current ownership of chickens or an intention to keep chickens within a two-year period. Few households were excluded based on the latter criterion. Two-stage sampling was used to give a total of 240 households in Sanza Ward and 280 in Majiri Ward: by first enrolling all eligible households with a child under 12 months of age, and then using random selection to enrol additional households with a child aged 12–24 months. Baseline data collection was completed for 229 households in Sanza Ward in May 2014 and 274 households in Majiri Ward in November 2014, as part of the staged implementation within the larger project design. An overview of administrative units within the study and the number of enrolled households at a village, ward and district level are shown in Fig 1. The timing of vaccination campaigns is shown. Newcastle disease (ND) control programs using the thermotolerant I-2 ND vaccine, administered via eyedrop [20] were established within the project sites, beginning in May 2014 in the first two of the eight villages (Fig 1). Local candidates for the role of “community vaccinator” were identified in consultation with village leaders. Training workshops were conducted using materials developed for and pre-tested in resource-poor settings [21]. These included both theoretical and practical components to cover aspects of chicken health and disease, principles of ND transmission and control, vaccine storage and handling, and logistical aspects of implementing community-wide vaccination campaigns. In addition to the six-month delay between commencement of research activities between the two wards, there was a staggered start to ND vaccination programs within wards. At ward-level meetings in both Sanza and Majiri, village representatives drew pieces of paper at random to designate their village for immediate or delayed introduction of ND vaccination, with a delay of three campaigns between the two groups (Fig 1). Prior to each campaign, community vaccinators visited households to register chicken numbers and establish chicken-keepers’ interest in fee-for-service vaccination, allowing an appropriate quantity of vaccine to be ordered. The cost of vaccination was initially set at 50 Tanzanian Shillings (TZS) per bird in all villages, and was increased to 100 TZS per bird in Majiri Ward in 2016 based on local consensus. The months of vaccination campaigns within the eight villages is shown in Fig 1. Studies on the epidemiology of ND in village chickens are limited, however general information from Veterinary Investigation Centres across different agroecological zones of Tanzania [22] suggests a heightened risk of ND outbreaks between July and November each year, during the dry season. Accordingly, initial recommendations were that vaccination campaigns be held in January, May and September. This timing was followed in Majiri Ward throughout the study. In Sanza Ward, vaccination months were changed to March, July and November in the second year of implementation. This transition was made both to reflect the perceived risk of outbreaks within the community, and to accommodate changes to the Tanzanian financial year in Tanzania and associated logistical challenges of distributing vaccine in January. Although it is considered prudent to conduct two ND vaccination campaigns prior to the high-risk period for disease outbreaks [23], delays in project inception led to the first campaign (in Chicheho and Ikasi villages) being held in late May 2014. This coincided with reports of illness and mortality in a small number of chickens within the area, suggesting the potential presence of a disease compatible with ND. A decision was made to proceed with the campaign, given it was considered a minor risk that chicken-keepers might attribute post-vaccination illness or mortality in their chickens either to the inefficacy of the vaccine or as a direct outcome of vaccination. In addition to hand-washing between each household visited, payment was postponed to a follow-up visit as an additional measure to reduce the potential for disease transmission, with the transfer of money identified as a potential pathway for viral transmission. Information used for this analysis falls into two broad categories: (1) relating to all households to which ND vaccination was made available, and (2) relating to the subset of households enrolled in the longitudinal study of children’s growth (hereafter referred to as “enrolled households”). In the former category, the total number of households was derived from initial census data collected by the project, and the number of households vaccinating their chickens in each campaign was determined from community vaccinators’ records. Daily rainfall data were recorded from a rain gauge with 1 mm graduations, located at the village office in Ikasi Village, Sanza Ward and Kinangali Village, Majiri Ward. Information on study participants was drawn from two questionnaires, collected through interviews by local enumerators recruited and trained within each ward. One questionnaire (applied at six-monthly intervals to mothers of enrolled children), focussed on maternal and child health and nutrition, while the other (applied annually to an intended equal number of male and female household members; actual sample comprised 60.5% female respondents of 1,354 completed questionnaires) encompassed demographic data, socioeconomic factors, livelihoods and chicken-keeping practices. Additionally, two representatives from each village (one male, one female) were employed as “Community Assistants” to collect ongoing data from enrolled households. The number of chickens owned, categorised by age (i.e. under or over two months) and participation in each vaccination campaign were recorded during twice-monthly household visits. Socioeconomic status of enrolled households was determined using a modified version of a “household domestic assets index” (HDAI), developed for use in sub-Saharan Africa [24]. The index assigns a weight to livestock and non-livestock assets according to their equivalent value. It is acknowledged that the relative value of assets will vary between settings, however the reference’s weighting system was considered adequately appropriate for the study sites to provide a reasonable estimate of enrolled households’ wealth. Given that some of the information involved in index construction (such as the size of land ownership and the age of assets) was not collected within this study, a modified index was formed based on livestock species and household items owned. Chicken numbers were excluded from these calculations, in order to evaluate their influence on vaccination uptake separately to their general contribution to household wealth. Language group, as a proxy for ethnicity and cultural practices, was also considered as a potential determinant of chicken ownership and vaccination uptake, based on observed and documented differences in household dynamics, diets and the practice of agropastoralism in this setting [25, 26]. Questionnaire responses for the “first language” of both the mother and father of children participating in the study were combined with information on the gender of the household head to determine the dominant language group of each household, likely to influence agricultural and dietary practices. The number of chickens owned by households was recorded at two-weekly intervals. While chickens of all ages were included in descriptive summaries, the number of chickens over two months of age was used as a predictor and outcome variable in inferential analyses. This distinction was made based on documented high rates of mortality amongst chicks in village settings from causes other than ND, such as predation, harsh weather conditions and poor nutrition [27, 28]. Percentages were determined for categorical variables and means and standard deviations or medians and interquartile ranges (IQR) calculated, for normally and non-normally distributed continuous variables, respectively. Records of chicken numbers in enrolled households were aggregated at village and ward levels for twice-monthly intervals over a two year period in both wards (beginning in July 2014 in Sanza Ward, and November 2014 in Majiri Ward). Graphical summaries were assembled for the percentage of enrolled households owning chickens and the average chicken flock size over time. In recognition of fluctuating levels of village chicken ownership, the percentage of households owning chickens consistently over a twelve-month period, and those owning chickens intermittently or not at all, were evaluated for each ward. Levels of vaccination uptake were determined for all households in a village, and amongst households enrolled in the study. In the former case, given the absence of information on chicken ownership across the whole village, vaccination levels were calculated relative to the total number of households recorded in the census (a proportion of which it is noted would not be keeping chickens). For the enrolled subset, households owning chickens were categorised as vaccinating or non-vaccinating according to their participation in each campaign. Depending on the type of response variable, linear mixed models (for quantitative variables) or generalised linear mixed models (for binary and count variables) were used, with the mixed model approach to allow for geographical clustering. For all analyses, univariable models were first used to test unconditional associations between predictor variables of interest (including socioeconomic and demographic characteristics and temporal factors) and the two outcomes of interest: chicken flock size and vaccination uptake. In the case of chicken numbers and asset scores, log-transformations were used to minimise the excessive influence of very large numbers. Variables with p-values under 0.1 were included in multivariable models, and a manual backwards elimination approach used with variables being retained if they were significant at the 5% level. All multivariable models included ward, village, subvillage and household identification as random effects, to allow for clustered data. Initial analyses explored: (a) longitudinal associations between enrolment in the study and participation in ND vaccination campaigns, and (b), for enrolled households, associations between the timing of commencement of vaccination (immediate or delayed) and levels of participation in the first campaign. The relationship between chicken numbers and ND vaccination was recognised as having the potential to be bi-directional: (1) with lower mortality amongst vaccinated birds resulting in increased chicken flock size, and / or (2) with households owning more chickens being more likely to invest in vaccination. Schematic diagrams of analyses conducted, including the relevant time frames to explore causality, are shown in Fig 2. All were within multivariable models. Vaccination in a given campaign was evaluated as a predictor of chicken flock size in the period following vaccination, by considering the mean number of chickens owned between one campaign and the next (Fig 2A; linear mixed model). To explore the alternative causal pathway, associations between chicken flock size and the decision to vaccinate were tested using the number of chickens owned at the time of vaccination (Fig 2B; generalised linear mixed model (binomial model)). Finally, the significance of participation in multiple vaccination campaigns over twelve months was tested as a predictor of the chicken flock size at the end of this period (Fig 2C; generalised linear mixed model (Poisson distribution)). The fit of the mixed models was assessed using standard residual diagnostic plot methodologies. All analyses were conducted using GenStat Release 18 (https://www.vsni.co.uk/).
DIMA AI Care
